Business Approach Comparison Sony PS4 and Microsoft Xbox

[liolio]

I agree with you Phil, playstation as a brand is still stronger the xbox in EU, it is more even in US.
There are other factors, ps4 merits aside: playstation is more opened: from the hardware to networking features. I think that as PS3 price went down last gen it showed, more budget conscious people saw through the extra cost of Xbox ecosystem. (/live), restriction on peripherals (/hdd), BRD player was a nice bonus too.

Now the ps4 benefits from the playstation brand's strength but I would say it benefits equally from Sony positioning toward a more opened platform. I think it feedbacks and in turn strengthen the playstation brand.
MSFT tried to go even further this gen on that front, with a system with lesser specs, sold at an higher price (though including a accessory) and it backslashes them badly> they have to actually open their ecosystem more than it was during the 360 era.
There are external factors too, more and more free services on Android/iOS, STB that does not require paywall, etc.

MSFT is back pedaling on all fronts, it is imo terrible. Imagine the score a student looking at a business case and coming with such a business plan would get... he would have gotten it all wrong and get a dreadful notes: bad analysis of what their user base wanted, bad analysis of how media consumption trend (focus on tv, put apps that are free elsewhere and used more and more behind a paywall), chose as primary focus selling points that were not encompassing with all their target territories (many different tv markets, voice recognition working well only with a couple of languages) and that are at best unproven with their core audience within those markets. And all that translates in "misplaced" R&D efforts (when OTS IP were providing all they should have ever needed, and ended up with a super complex developing environment /3 OS).

Looking at what the new management is doing, I think they are really few chances that they would have greenlighted the project.
Looking back in time, the yukon platform was a great basis, especially the price they were aiming at (225$): that was their best bet to expand the market, no system offering sane core gaming capabilities launched at such a price in a long long while.

What they are doing now is perfect or close, they are reacting really swiftly. Though without a consistent subsidizing effort I can't see them ending up with more than a big third of the global market.

 

[milk]

Color me exited when they announce a PowerBrickless ONE. The thing's perpetually active cooling fan (what power-brick needs its own fan for gods sake?) made the instant resume a non-feature since I had to turn that option off to get some quiet in my home. I wanted the ps4, but bought the ONE so my family could play great Kinect games since they talked so much about freaking Just Dance and that sort of stuff. In a act of complete selflessness I pretended to not read digital foundry and beyond3d and bought the weaker ps4, last week, just to discover the thing will probably not get that many great Kinect games anymore.
I forgot the golden rule: Never be an early adopter.

 

[Shifty Geezer]

I thought Just Dance was uber compelling. If MS had really evolved that side of their console, they may have gotten me to care. At the moment both boxes are just generic game boxes and neither represents a must-have experience. Potentially, PS4 may become an upgrade box when it's cheap and has a broad library of nonsense I care for, but I've zero plans yet.

 

[Arwin]

The problem with Kinect is that the audience for it buys consoles at 249, not 499. When the original Kinect launched, they weren't that far off that target. But right now, they're very far off and don't have much hope to catch up, when right now it hurts early adoption among the more hardcore crowd who don't care about motion gaming and feel they have to pay a lot of money for a 'lesser' box.

 

[Shifty Geezer]

That's probably quite true. Perhaps MS should have left Kinect off the cards until they could do a different box with cheaper hardware, and just stuck with the console to replace XB360 for now? That seems to be what they're moving back to.

 

[MrFox]

The thing's perpetually active cooling fan (what power-brick needs its own fan for gods sake?) made the instant resume a non-feature since I had to turn that option off to get some quiet in my home.

The very cheap power bricks need fans, because they are inefficient, they get hot, and passive cooling would have been a few dollars more. This tiny fan will eventually fail with dust build up. And even if you clean it regularly, they usually have an MTBF around 25,000 hours. That's 3 years if you leave it on 24/7.

There are people who are trying to buy a spare XB1 Power Supply because they don't want to have down time waiting for a replacement. MS refused to sell it to them, they only sell it if you send back a faulty one.
http://www.reddit.com/r/xboxone/comments/1vajjq/a_guide_on_how_to_order_a_spare_xbox_one_power/

MS should stop trying to make consumer electronics and just outsource the integration job to a firm that knows what they are doing.

 

[liolio]

That's probably quite true. Perhaps MS should have left Kinect off the cards until they could do a different box with cheaper hardware, and just stuck with the console to replace XB360 for now? That seems to be what they're moving back to.

It is a given. Actually pushing that hard to have it on the xb1 was close to crazy when their own efforts to keep kinect game library afloat was at best limited, one understatement a day doesn't hurt I guess.

For me the xb1 is somehow part of MSFT "past" as it happened before Balmer initiated a significant reorg of the company and let new people take the lead.
Higher executive might have known significant changes were coming and now that is all said and done I think they should have delay the launch, either way it should have launched earlier. I know tech they chose was not ready but it does not make the timing any better. Actually the tech was not completely ready at launch either, aka Jaguar power management features do not work, it quite a bit of a shame that both system pulls ~40 Watts while navigating the menus (iirc).

For me it is a bad product, MSFT inner troubles made it that way. It is not a matter of price, if I buy something "MSFT" in the upcoming years it is more likely to be a gaming enable HTPC running direct x12, I don't care if it is powered by x64 or ARM cpus. Somehow MSFT still has a shot with me.

Where things get weird is that me doing so is a greater win for MSFT than me buying a XB1, that is the core of the problem and why the xb1 doesn't blend that well in the new priority set at MSFT HQ ( that is my pov and only mine) and why the overall development looks confused.

 

[Scott_Arm]

http://arstechnica.com/gaming/2014/...decouple-kinect-from-xbox-one-began-in-april/

Phil Spencer seems like a smart guy. He basically admits the customer is right and didn't waste any time making changes. You can attack the company as a whole for not sticking with the plan, but in the end this guy seems to be making smart business decisions without any BS. I wanted the original vision, but if it doesn't sell it doesn't sell. Things are early enough along that they can improve the product and make it desirable for consumers. Sticking with the original plan would have just been ego driving the company into the ground. That's one good thing that can happen from switching management. Sometimes people create a project and they just can't let it go because their career is riding on it. You bring someone else to the top, and it's not their baby, and they can do the right thing.

 

[wco81]

http://arstechnica.com/gaming/2014/...decouple-kinect-from-xbox-one-began-in-april/

Phil Spencer seems like a smart guy. He basically admits the customer is right and didn't waste any time making changes. You can attack the company as a whole for not sticking with the plan, but in the end this guy seems to be making smart business decisions without any BS. I wanted the original vision, but if it doesn't sell it doesn't sell. Things are early enough along that they can improve the product and make it desirable for consumers. Sticking with the original plan would have just been ego driving the company into the ground. That's one good thing that can happen from switching management. Sometimes people create a project and they just can't let it go because their career is riding on it. You bring someone else to the top, and it's not their baby, and they can do the right thing.

It's great to be adaptable but can they overcome some of the initial design decisions?

They clearly sacrificed some of the performance capabilities to fit the Kinect into the BOM.

Now they've removed the Kinect but the chipset they chose to accommodate the Kinect still remains.

 

[joker454]

For me the xb1 is somehow part of MSFT "past" as it happened before Balmer initiated a significant reorg of the company and let new people take the lead.

It really does seem like their has been a changing of the guard and focus has dramatically shifted. Now they seem to be all about multiplatform cloud and services, basically getting people hooked on smaller frequently recurring fees on whatever platform they own rather than one off larger fees every now and then on only the platforms they sell. Fortunately the xb1's hardware is generic enough that it can be software molded to whatever future they have in mind for it, although I get the feeling that it may take the better of a year until we see where things go.

 

[Scott_Arm]

It's great to be adaptable but can they overcome some of the initial design decisions?

They clearly sacrificed some of the performance capabilities to fit the Kinect into the BOM.

Now they've removed the Kinect but the chipset they chose to accommodate the Kinect still remains.

What chipset to accomodate Kinect? Are you referring to the CPU/GPU?

 

[wco81]

Yeah, could they have used the same CPU/GPU as the PS4 or even better?

Use the same RAM?

 

[Scott_Arm]

Yeah, could they have used the same CPU/GPU as the PS4 or even better?

Use the same RAM?

Who knows. Sony was betting on GDDR and MS seemed to believe it would be unaffordable at 8GB. Hindsight is 20/20.

 

[Rangers]

From Bkilian's posts DDR3 and ESRAM was the plan for Xbox from the start. In fact, the change to 8GB was made late in the game (late 2011). Before then, they were planning a 4GB DDR3/ESRAM setup.

MS never considered GDDR5 more or less, I think. They are really hung up on the whole ESRAM/EDRAM concept. They started with ESRAM and built from there.

Now, it's possible they somehow built from wanting to be really cheap in order to fit Kinect in the BOM, and that necessitated the DDR3 which in turn necessitated the ESRAM, but I really dont even necessarily get that feeling. I have a feeling "cheap" came in the decision to go with less CU's.

 

[bkilian]

From Bkilian's posts DDR3 and ESRAM was the plan for Xbox from the start. In fact, the change to 8GB was made late in the game (late 2011). Before then, they were planning a 4GB DDR3/ESRAM setup.

MS never considered GDDR5 more or less, I think. They are really hung up on the whole ESRAM/EDRAM concept. They started with ESRAM and built from there.

Now, it's possible they somehow built from wanting to be really cheap in order to fit Kinect in the BOM, and that necessitated the DDR3 which in turn necessitated the ESRAM, but I really dont even necessarily get that feeling. I have a feeling "cheap" came in the decision to go with less CU's.

I suspect cheap came up in most of the decisions. RAM prices stay relatively the same or drop slowly, process reduction can cause much larger decrease in cost per chip. So maybe they went with eSRAM because it will share in the process cost reduction. I don't know the reasons. I do know eSRAM was pretty much always in the plan.

 

[liolio]

Yeah, could they have used the same CPU/GPU as the PS4 or even better?

Use the same RAM?

I find it interesting to look back at that "Yukon" platform and to compare it to the shipping XB1:
*The ALUs count for the gpu is "strange" so I ignore it for now.
*The performances goals for the platform was reasonable: 8 times the 360.
*They wanted the system to be scalable.
*They were clearly depicted a 8 cores system.
*They considered both 28nm and 22 nm SOI.
*They were considering 4GB of RAM on a 128 bit bus, DDR4: they were clearly aiming at +50GB/s to the main ram.
*They wanted lots of eDRAM, 32MB being the bottom figure.
*They wanted BC.
*They wanted a SoC.

The most interesting part for me is the price for the SOC and the system: 50$ and 225$.
Power figures are also interesting 50 Watts for the SoC, 120 Watts for the system.
Overall it is pretty good at depicting what the xb1 was to be, it shows that Kinect acceleration was part of the design.

Where it gets interesting is how they adapt the system to technological advance and business conditions.

First thing first that 22nm SOI or not process was not an option.
Actually iirc best IBM processors still use 32nm process. Blue gene Q is using 45nm.
Putting business consideration aside, it turned out that 22nm and lots of edram was out of the picture. For unknown reasons, IBM 32nm process was discarded too. Everybody bet on price and other business considerations,so do I though it is just a bet.
So they went with a 28nm process, no Edram. At same the time BC had to go.

It turned out that DDR4 got repeatedly pushed back. They could not have their +50GB/s of bandwidth to the main RAM.


That is where it gets interesting, BC could not be delivered, bandwidth was an issue, so was the scratchpad, yet it seems they did not go back to the drawing board. More on that later.

To their credit they did manage to make it, they managed to get 32MB of on die memory, they brute forced their bandwidth issue by doubling the bus size.
So quite an achievement? Well as I just said they made it. Yop they got what they want as far as the silicon is concerned but their one thing they got very wrong: the BOM
They were aiming at 225$ for the system without Kinect. IHS estimate may not be too far off as it is in line with MSFT own claims. Doubling the RAM is about 30$.
Kinect BOM may also have ended a bit more than expect as the accelerator in charge skeletal tracking did not end up on the SoC as shape but in Kinect to lower the amount of data shared between the console and the sensor, though nothing to throw the project unbalanced that much.

It gets really interesting, even with an estimate cost of 75$ for kinect, the overall BOM from Yukon + kinect would have 300$. Let me rephrase it : 300$.
I should not make a big thing out of it as the slides stated a 299$ price tag.
Something is not OK here.

At this point it is interesting to reconsider other data present on the Yukon platform slide, the SoC power consumption: 50 Watts. Even more interesting the SoC price 50$.

How did they hope to have XB1 level of performance for that power with CPU running at 2GHz and a GPU running at 1GHz? Even if they are just ballpark figures?
How did they hope to have that amount of processing power for 50$, that is in their best case scenario using a 22nm SOI process allowing for eDRAM?
MSFT engineers knows better than that, they can't be that far off.
I think they were not off at all, project Yukon was simply not meant to have the level of performance the Xbox One delivers.

With that in mind, as well as MSFT "literature" ie multiple interviews showcasing their views on technology, GPU, GPU efficiency, etc. I think that the figures they gave for the GPU makes sense, a lot of sense.
Project Yukon was meant to embark only 4CU/SIMD, 64 ALUs, an ALU being a vliw5 or vliw4 block, there are 16 such block per SIMD in old AMD GPU (still be there in GCN with few changes).
MSFT own word, they vouch Xenos efficiency at 60%, Vliw 5 improved on that, so did VLIW4 and GCN improve further.
Let make a gross calculation, on one side 48 ALU (xenos) at 500MHz with 60% efficiency on the other 64 at 1GHz with 90% efficiency.That is a 4x performance on ALU alone.
Add in improvement to texture units, ROPs, the bigger edram, 8 time the RAM, more bandwidth to the RAM, 6 OoO CPU dedicated to games, 2 for the OS and services, the accelerator for Kinects, actually I would not be surprised if an engineer depict such a system a ~8 times as potent as the 360. FLOPS are a thing for forum and customers not really engineers.

To my point, I think that for a long while MSFT plan was to offer it customers more of an upgrade path than a powerhouse of a console.
MSFT might do lots of market studies, they might be aware of their brand deficit against playstation. They may have also know OS were to become prevalent everywhere including the living room.
They did not want to unbind customers from their ecosystem by breaking BC. Even more than Nintendo MSFT knows that legacy support is incredibly important, so are locked in customers as a less pleasant way to state it.
BC doesn't sell a system but lack of BC impact it negatively and put your user base at risk.

Hell broke loose, whatever the reason it is irrelevant to my point, back compatibility could not make it into the system, damned!
All those customers would be on the market again so to speak, free to choose the system they wanted.
The affordable upgrade strategy fell apart, shit happens the wise man says. What would have happen if such a system shipped? When was that decision taken? I don't care, reading BKillian it seems it could have been a good long while ago. Point is they needed a more potent system and there are also timeline considerations, the system could have been pushed back in the face of the 360 running success or other business considerations.

That is where I've been heading with this long post and specifically my "more on that later", what is interesting is what they did:
They still made it, they made project Yukon happen:
it is all there, 32MB of on die memory, the accelerators, the +50GB/s of RAM between the SoC and main RAM, the 4GB of RAM.
The 32MB costed them lot of silicon, to achieved the bandwidth they wanted they doubled the bus width.
To have the performance boost they wanted they scaled the GPU significantly, they also double the amount of memory.

They made it and that is the issue I think before leaving Balmer wanted to address and the new management too. It looks like MSFT being that big some decisions can't be changed, some size and cultural issue within the company, too long heavy decision process which make people unwilling to question validated decisions once they jumped through all those damned many hoops and got finally validated, etc.

Project Yukon had been "decided", once that was done... hell could break loose it still had to happen.
You could have scaled it up, accomodate it, etc. which they did, but there was not going back.
They had to make more out of project Yukon, I think they never went back to the drawing board.

Scaling up project Yukon ended incredibly costly, especially looking at what seemed to be MSFT performances target and to the despair of all their fans they were lower than Sony's ones.

So my incredibly long answer about "what they could have used?", is more about what they could have done:
they should have gone back to the drawing board and design from scratch a system that fits their performances targets instead of scaling something that should have been discarded altogether.
Kinect inclusion in the std package should have been reavaluated once the lowest BOM they could get for their perfs figures would have been determined. I think their primary plan wrt Kinect, tv, services required a price as low as possible. I think BC was important to lower the risk of user base changing side.

I'm not an engineer but looking at Yukon and at Durango I think this is the hardware MSFT should have come with (/same perf targets and a 2013 release):
System:
2 jaguar cluster, 1 core is disabled, 1.6GHz
12 CU, 1 disabled, 850MHz
16 ROPs
one MEC
one "geometry" engine.
True audio.
192 bit bus.
6GB of GDDR5 +110GB/s bandwidth to the main RAM (try to use relatively cheap gddr5).
2GB reserved for the OS.
below 250mm2.

Kinect:
I think it is either too cheap or too expensive. I would not have included it but who care about my pov?
I think they should in fact have been a bit more ambitious:
Instead of using some custom silicon like SHAPE and that "skeletal" accelerator they should have considered a software solution running on an OTS ARM SoC (included in kinect).
It should be standalone, powered by USB and enable to communicate via WIFI with any windows device or the xbox.
The BOM could have been closer to 99$, standalone selling around 120$.
It would have offer the same functionality to the One but also other Microsoft products, processing would have been made completely on Kinect, sports/fitness games could have been ported to people PC (laptop and tablets even phones).

Anyway reading Bkillian's comment make me think that XB1 design choice are based on something designed too long ago and that did not get "reviewed" again even in the face of pretty significant strategic change(s).
I guess that is why they wanted scalable to begin with: so nobody has to go the painful validation/decision process again, quite scary when you think of it.

 

[liolio]

I want to add that it looks like a completely different process than at Sony HQ, whereas the PS4 is clearly not a single man work there was definitely only one guy in charge.
So I would not say that Cerny is better than MSFT guys, may be he is, but I think that MSFT engineers were not anywhere near as "in charge" as he was.

Durango is something designed in committee, a design whose basis have been settled upon too long ago.

 

[Deleted member 86764]

Microsoft did so much right with the 360, it's a shame that they've failed so spectacularly this time around. Not just with the sales but with the hardware and the message too.

 

[Deleted member 86764]

Durango is something designed in committee, a design whose basis have been settled upon too long ago.

This reminds me of an old saying; a camel is a horse designed by committee. You could well be right with this.

 

[(((interference)))]

Well, you know their original price target was XB1 with Kinect for $399.